**3. Results**

#### *3.1. Analysis of Gut Microbiota, Microbiome and Fecal Metabolome During NEC-1*

To understand the microbial and metabolomic evolution during the early onset of necrotizing enterocolitis (NEC), we studied clinical profile suspected NEC (NEC-1) preterm infants. NEC-1 children underwent more glycopeptides treatment, showed significantly higher cordon lactates, bacteremia and a longer full enteral feeding, when compared to age-matched healthy children (Table 1). NEC-1 children also displayed a lower plasma pH and enteral milk volume at day 7 (Supplementary Figure S1A,B) and a higher abundance of *Streptoccoccus* species (Supplementary Figure S2A) compared to healthy children. Both populations of children showed a high intragroup variance in terms of gu<sup>t</sup> microbiota (Supplementary Figure S2B) and overall microbial diversity (Supplementary Figure S2C). NEC-1 microbiome showed increased activity for pathway related to transcription, glycosaminoglycan degradation and lysosome, compared to healthy children (Supplementary Figure S2D). Then, we analyzed the fecal metabolome to appreciate NEC-1-induced changes in gu<sup>t</sup> microbial metabolic activity. NEC-1 children displayed a reduced intragroup variation and significantly lower levels of ethanol (Supplementary Figure S2E). Overall, these data show that NEC-1 is characterized by a precise gu<sup>t</sup> microbiota, microbiome and gu<sup>t</sup> microbial metabolites profile.

#### *3.2. Analysis of Gut Microbiota, Microbiome and Fecal Metabolome During the Evolution of NEC-1 over the First Two Months of Life*

Given the presence of a NEC-1-specific gu<sup>t</sup> microbiota and microbiome profile, we aimed to identify at what time these profiles establish. We divided both NEC-1 and healthy children populations in subgroups according to periods of ten days of life as it follows: 1–10 d (d stands for "days"), 11–20 d, 21–30 d for the first month of life and > 30 d for the second one. In the first 10 days, NEC-1 children displayed a divergent and more homogenous gu<sup>t</sup> microbiota compared to healthy children, with the latter characterized by a higher abundance of *Klebsiella* species (Figure 1A,B). At this stage of life, gu<sup>t</sup> microbiota in NEC-1 had a lower diversity based on Chao-1 index (Figure 1C) and a di fferent microbial activity related to replication, recombination and repair proteins, lysosome and glycosaminoglycan degradation (Figure 1D). No significant changes were observed in fecal metabolites (Figure 1E). Overall, these data show that gu<sup>t</sup> microbiome starts to diverge at the early onset of NEC-1.

**Figure 1.** Analysis of gu<sup>t</sup> microbiota, microbiome and metabolome in the first 10-days of life in healthy vs. necrotizing enterocolitis (NEC)-1 children. ( **A**) Gut microbiota analysis via linear discriminant analysis (LDA) score between healthy (H) vs. NEC-1 children, in the first 10-days of life, 1 to 10 days; (**B**) principal component analysis (PCA) of the gu<sup>t</sup> microbiota; ( **C**) indices of gu<sup>t</sup> microbiota diversity; (**D**) LDA score for microbial pathways; (**E**) histogram of the overall fecal metabolites and PCA as inset. \*\**P* < 0.01. Two-way ANOVA, followed by a two-stage linear step-up procedure of Benjamini, Krieger and Yekutieli to correct for multiple comparisons, by controlling the false discovery rate (<0.05); *N* = 15 for H and *N* = 4 for NEC-1.

In the second 10-days of life, NEC-1 gu<sup>t</sup> microbiota was characterized again by a higher abundance of *Streptococcus* species and bacteria from the Micrococcales order (Figure 2A), with a high intragroup variance (Figure 2B). At this stage of life, NEC-1 gu<sup>t</sup> microbiota also showed a higher diversity based on Chao-1 index (Figure 2C), but no microbial pathway di fferently regulated (Figure 2D). As for the fecal metabolome, NEC-1 children displayed significant lower levels of serine (Figure 2E). Overall, these data show a stronger evolution of gu<sup>t</sup> microbiota than gu<sup>t</sup> microbiome in the second 10-days of life, between NEC-1 and healthy children.

**Figure 2.** Analysis of gu<sup>t</sup> microbiota, microbiome and metabolome in the second 10-days of life in healthy vs. NEC-1 children. (**A**) Gut microbiota analysis via LDA score between healthy (H) vs. NEC-1 children, in the second 10-days of life, 11 to 20 days (d) (the LDA score is only shown for NEC-1 children meaning that no bacteria are significantly higher in the H group vs. NEC-1); (**B**) PCA of the gu<sup>t</sup> microbiota; (**C**) indices of gu<sup>t</sup> microbiota diversity; (**D**) null cladogram for microbial pathways; (**E**) histogram of the overall fecal metabolites and PCA as inset. \*\**P* < 0.01. \*\*\**P* < 0.001. Two-way ANOVA, followed by a two-stage linear step-up procedure of Benjamini, Krieger and Yekutieli to correct for multiple comparisons by controlling the false discovery rate (<0.05); *N* = 14 for H and *N* = 10 for NEC-1.

In the third 10-days of life, changes in NEC-1 gu<sup>t</sup> microbiota compared to healthy children occurred to a bigger extent and were related to increased *Staphylococcus* and *Streptococcus* species (Figure 3A,B), together with a high intragroup variance (Figure 3C) and no change in the overall diversity indices (Figure 3D). We also observed a NEC-1 microbiome profile, mainly based on thiamine and seleno-compound metabolism (Figure 3E). The NEC-1 gu<sup>t</sup> microbiota profile of the third 10-days of life was associated with: (i) multiple diseases and found significantly increased in ulcerative colitis (Figure 4A); (ii) host genetic variation and significantly related to ANP32E, a gene involved in ulcerative colitis [22], in line with previous reports. In terms of fecal metabolome, we observed no significant changes in NEC-1 vs. healthy children (Figure 4C). Then, we studied feces collected in the second month of life. In this period of life, the taxonomical differences in the gu<sup>t</sup> microbiota of NEC-1 vs. healthy children were related to the increase in *Raoultella* species in NEC-1 gu<sup>t</sup> microbiota (Figure 5A), with a still high intragroup variance (Figure 5B), and no change in the overall microbial diversity indices (Figure 5C). We also observed microbial functions related to DNA repair increased in the NEC-1

gu<sup>t</sup> microbiome (Figure 5D). This period of life was characterized by the highest separation in terms of fecal metabolome, with significant lower levels of ethanol and leucine in NEC-1 children (Figure 5E).

**Figure 3.** A specific gu<sup>t</sup> microbiota and microbiome exist in the third 10-days of life in healthy vs. NEC-1 children. (**A**) Comparative analysis of the gu<sup>t</sup> microbiota by LDA effect size (LEfSe): the cladogram shows bacterial taxa significantly higher in the group of children of the same color, in the fecal microbiota between healthy (H) vs. NEC-1 children, in the third 10-days of life, 21 to 30 days (**D**) (the cladogram shows the taxonomic levels represented by rings with phyla at the innermost and genera at the outermost ring and each circle is a bacterial member within that level); (**B**) LDA score used to build the cladogram in (A); (**C**) PCA of the gu<sup>t</sup> microbiota; (**D**) indices of gu<sup>t</sup> microbiota diversity; (**E**) LDA score for microbial pathways. *N* = 13 for H and *N* = 7 for NEC-1.

**Figure 4.** Diseases, host genetic variation and metabolome analysis in the third 10-days of life during NEC-1. (**A**) Diseases and (**B**) host genetic variation linked to NEC-1\_21–30d associated gu<sup>t</sup> microbiota; (**C**) histogram of the overall fecal metabolites and PCA, as inset. *N* = 13 for H and *N* = 7 for NEC-1.

**Figure 5.** Analysis of gu<sup>t</sup> microbiota, microbiome and metabolome in the second month of life in healthy vs. NEC-1 children. (**A**) Gut microbiota analysis via LDA score between healthy (H) vs. NEC-1 children, in the second month of life > 30 days (d); (**B**) principal component analysis (PCA) of the gu<sup>t</sup> microbiota; (**C**) indices of gu<sup>t</sup> microbiota diversity; (**D**) LDA score for predictive microbial pathways (*P* < 0.01); (**E**) histogram of the overall fecal metabolites and PCA, as inset. \*\*\**P* < 0.001. Two-way ANOVA, followed by a two-stage linear step-up procedure of Benjamini, Krieger and Yekutiel,i to correct for multiple comparisons by controlling the false discovery rate (<0.05); *N* = 11 for H and *N* = 6 for NEC-1.

#### *3.3. Specific Impact of NEC-1 on the Evolution of Gut Microbiota. Microbiome and Fecal Metabolome over the First Two Months of Life, Compared to Healthy Children*

To investigate the evolution of gu<sup>t</sup> microbiota, microbiome and fecal metabolome over the first two months of life, we conducted an intra-group study in both NEC-1 and healthy children, according to the four groups reported above: 1–10 d, 11–20 d, 21–30 d and > 30 d. We did not observe any taxonomic significant change in the gu<sup>t</sup> microbiota of NEC-1 children. However, the group NEC-1\_21–30 d had a specific gu<sup>t</sup> microbiome with an increased restriction enzyme activity, among others (Supplementary Figure S3A). The four NEC-1 groups also differed in terms of fecal metabolome, with regard to leucine, ethanol and serine amounts (Supplementary Figure S3B). Based on these results, we performed a metabolomic enrichment analysis on two levels: (i) pathway-associated metabolite sets (Supplementary Figure S3C) and (ii) single nucleotide polymorphism (SNP)-associated metabolite sets (Supplementary Figure S3D). NEC-1 metabolomic profile (increased ethanol and serine) was significantly associated with both homocysteine degradation and phosphatidylethanolamine biosynthesis (Supplementary Figure S3C), with serine being the metabolite the most linked to NEC-1-associated SNP (Supplementary Figure S3D). By contrast, in healthy children, the four groups reported above did not differ in terms of both gu<sup>t</sup> microbiota and microbiome, but only with regard to fecal metabolome (Supplementary Figure S4A). Healthy metabolomic profile (increased leucine, ethanol and dihydroxyacetone) was significantly associated with valine, leucine and isoleucine degradation and to ketone body metabolism (Supplementary Figure S4B), with leucine being the metabolite the most linked to healthy-associated SNP (Supplementary Figure S4C). Overall, these data sugges<sup>t</sup> that: (i) a different intragroup evolution exists between NEC-1 and healthy children, with regard to gu<sup>t</sup> microbiota and microbiome; and (ii) the NEC-1 microbiome appears to be more sensitive to mother-related factors.

#### *3.4. Maternal and Child Factors Influencing the Gut Microbiota, Microbiome and Fecal Metabolome During NEC-1*

Next, we asked which factor related to both mother and child may a ffect the most the above reported parameters. We analyzed six conditions: neonatal antibiotherapy (ABx), ABx treatment on mother, childbirth (C-section (C-sec) vs. vaginal birth (VB)), very low birth weight (VLBW), extreme low birth weight (ELBW) and gestational age (GA) > or ≤ 28 weeks.

Only neonatal ABx treatment a ffected the gu<sup>t</sup> microbiota in both NEC-1 and healthy children (Supplementary Figure S5A). By contrast, all the above factors, except the VLBW, a ffected the gu<sup>t</sup> microbiome (Supplementary Figure S5B–F). Note that childbirth modality, ELBW and GA a ffected the gu<sup>t</sup> microbiome only in NEC-1 children (Supplementary Figure S5D–F). Moreover, all the above factors, except the neonatal ABx treatment and ELBW, a ffected the fecal metabolome between NEC-1 and healthy children (Supplementary Figure S6A–F). Then, we performed a metabolomic enrichment analysis on the pathway-associated metabolite sets, based on Supplementary Figure S6F, in which there is an increase in ethanol and succinate within in the NEC-1\_GA ≤ 28 w. Ketone body and butyrate metabolism were the most significantly associated with this metabolomic set (Supplementary Figure S6G).
